Sea-ice mapping and digitization in Antarctic waters using a shipborne monocular camera

An automated process is proposed for sea-ice field mapping and digitization using a monocular, machine vision camera. The proposed process for 2D mapping involves a number of steps, from image rectification and estimation of a virtual birds-eye-view to image segmentation by means of three distinct methods, both of individual images and large 2D maps. Lastly, the resulting ice floes are expressed as either convex or concave 2D polygons. Additionally, a separate process for 3D mapping is introduced, termed vSLAM. The implemented methods enable an automated and continuous acquisition of a detailed map of the ice field, as a series of 2D polygons and even 3D point-clouds, at engineering scale. Compared to other methods using satellite or aerial data, the proposed method is a more cost-effective and easy to integrate solution into vessels traveling in ice covered waters. The information can be highly compressed, thus enabling a higher level of situational awareness not only for the data collecting vessels, but also for others traveling in the same areas.